Printing quality: pursuing perfection

Hi Smartrappers,
so here is one more post about the pursue of perfection in printing quality.

In short, here are the two problems that I tried to address about the (outdated) design of the Smartrap I have: Z shaking/wobble and X end shaking. I know for sure that both have been addressed in the current Smartrap design, and I am looking forward to reading the first feedbacks.

My current results: the X end and Z axis are much more stable, wobble is reduced but still visible.
Drawbacks: with mechanical endstops I cannot use auto-bed leveling (but I was not using it anyway...).

Open problems: horizontal inaccuracies are still visible, either because of Z wobble or other source: maybe GT2 belts are not super-hyper precise? But I cannot tighten them further, otherwise the carriages become too hard to move and motors miss steps.

I get the best print quality around 0.1mm layer width, while the more I go below this threshold the more I get many tiny wires that ruin the surface, although extrusion is not a problem by itself even at 0.03mm (all this with Reg's Genie hotend). I have not tried less than that. My current guess is that it is a problem of orthogonality: the nozzle is slightly sloping, so it creates these tiny wires in particular in one direction.

Things I would like to try, besides some kind of touchless endstop:

• more stable wooden structure to further reduce wobble
• design of an adjustable head to see if orthogonality is the issue below 0.1mm.
• more symmetric design of GT2 belts mount, or maybe fishing line?

Any comments / suggestions / ideas?

Here are some pictures about the way I modified the original (again, outdated with respect to Serge's lastest version) design of the Smartrap I have.


Wooden structure, with adjustable joints for the Z smooth rods. They are significantly more stable.


Longer joint between Z motor shaft and threaded rod, fixed with screw and nut at the bottom. It also uses 3 nuts instead of two on the threaded rod. It reduced a little Z wobble, since the threaded rod is more constrained and more in line with the motor shaft.


X end for Reg's Genie hotend, without moving parts. X rods are also fixed with screw and nut. The fan is held by aluminium wire.
The hotend is definitely more stable.


My favourite vase: on the left, my best result some weeks ago: 0.2mm layer height, looot of Z wobble and other shaking issues. On the right my current best result obtained after applying the above changes to the original Smartrap design: 0.1mm layer height, Z wobble barely (if at all) visible in the picture but still present, tiny plastic wires here and there (but many on the bottom left, creating a sort of incrustation). The vase is about 6cm tall, sorry for low quality of the picture.

What are your best results?

Looks like some seriously impressive print quality. So why do you say that you can't use auto leveling with mechanical endstops?

Quote
knife-edge
So why do you say that you can't use auto leveling with mechanical endstops?

Well, I had to remove the joint in the head that allowed the hotend to rise and trigger the endstop when the bed was hit. Now the endstop works only in a corner during the homing phase, thanks to a little cube that I added close to (0,0,0). This is because I don't have any servo here.
But even if I had it, the autolevel procedure never worked properly for me, so I abandoned it for the moment.

Here you can see the small cube I put near the Z endstop, close to (0, 0, 0):

Here is some quite extreme test. In the picture there are 4 cubic vases 10x10x10 mm, printed with only 1 perimeter and no infill: the easiest test.
From left to right, they are printed at 0.05, 0.03, 0.02 and 0.01 mm layer height. Layers would be invisible in the picture (and even at naked eye), if it wasn't for the Z wobble (mostly). Z wobble becomes more and more visible by decreasing layer height, and at the same time the incrustations increase until they totally ruin the piece at 0.01 mm layer height.

So the good news is that printing at 0.01 is probably possible (thanks again Reg for this magic hotend!), but some improvements are necessary and it is not clear to me which ones. Orthogonality of the nozzle tip? Better cooling of the piece?

However, with this Z wobble (and probably other kinds of inaccuracy due to belts or I don't know what) printing at 0.01 is useless, because those damn waves are 0.8 mm, that is 80 times taller!



Edit 1: as a final note, even if Z wobble was resolved, horizontal resolution is the limit. Even with a 0.3 nozzle, printing at a layer width lower than 0.3 does not give nice results, so the nozzle diameter is the horizontal resolution. There are people printing with 0.1 nozzles, although I really wonder how they can clean them after a clog...
But even without going to such an extreme, having a 0.2 nozzle would be great already. Looking forward to it, Reg! Meanwhile, I will try to get rid of this Z wobble...

Edit 2: I forgot to mention that the test was done with 0.5 mm nozzle (and 0.7 mm layer width).

Edited 2 time(s). Last edit at 11/01/2014 09:57PM by cristian.

Your sub-0.1mm layer height prints are impressive. You must have your extruder finely tuned to compensate for that much compression?

Hi Cristian,

The wobble, is it in your case caused by an off-center ax coupler, or is the rod itself a bit bend? (or a bit of both?)

Quote
runninfarmer
You must have your extruder finely tuned to compensate for that much compression?

What do you mean exactly? There is no more compression than printing at 0.1 or 0.3 (as long as we neglect that layer width was 0.7 and the nozzle only 0.5). Moreover, I am printing with the standard Smartrap filament feeder which has no gears...
My only exceptional equipment with respect to the extruder of a standard Smartrap is Reg's Genie hotend.

Quote
BackEMF
The wobble, is it in your case caused by an off-center ax coupler, or is the rod itself a bit bend? (or a bit of both?)

Very likely a bit of both, and for sure an off-center ax coupler. I may try to print it again with a bit more care about orthogonality...

With a 0.5mm extruded diameter, if you're extruding rate is the same at 0.1mm vs 0.3mm layer height, how would you not have different layer widths with out decreasing extrusion rate? You'd have more compression at same rates for 0.1mm layer height, unless I'm wrong?

Quote
runninfarmer
With a 0.5mm extruded diameter, if you're extruding rate is the same at 0.1mm vs 0.3mm layer height, how would you not have different layer widths with out decreasing extrusion rate? You'd have more compression at same rates for 0.1mm layer height, unless I'm wrong?

As far as I know, the slicer takes care of all this automatically. Moreover, in the particular case of 1-perimeter vases, if the extrusion rate was not perfectly tuned the perimeter width would not be exact but it would not change much in terms of pressure.

Hi Cristian,
Nice results from your machine, regardless of the Z wobble.

Out of interest sake, which slicing program are you using? To me it looks like you may be using Cura??
If so, can you try using Slic3r and see if that changes anything in the results?

---

[regpye.com.au]
"Experience is the mother of all knowledge." --Leonardo da Vinci

Quote
BackEMF
The wobble, is it in your case caused by an off-center ax coupler, or is the rod itself a bit bend? (or a bit of both?)

Very likely a bit of both, and for sure an off-center ax coupler. I may try to print it again with a bit more care about orthogonality...[/quote]

I'm having similair wobble issues on my new build. and it seems also be a combination of the two.
For now i managed to compensate the coupler wobble by turning the rod so that the two wobbles cancel eachother out a bit (it still wobbles a bit)
I went to 2 hardware stores today only to find rods that are uncapable of rolling on a flat surface. :-) Not bend a lot, but bend enough to leave them on the shelf and look further..

Quote
regpye
Out of interest sake, which slicing program are you using? To me it looks like you may be using Cura??
If so, can you try using Slic3r and see if that changes anything in the results?

At the very beginning, when I first got my Smartrap, I started with Slic3r but results were poor. So I switched to Cura, with much better outcome. However, after some weeks, I started to feel that Cura did not give me all the freedom I needed, so I went back to Slic3r and stuck to it.

Cura gives better results than Slic3r in some cases (the slicing is smarter) but is quite limited. With some care, Slic3r can often outperform Cura, although it suffers from several weird issues in the slicing sometimes...

In any case, for these particular prints (spiral vases) the slicer is not important because the only free variable is the average flow. Besides it, there is nothing else to do for the printer than making the nozzle turn around and around in an endless continuous coil (no jumps, no retraction...).

After looking at the printing process for some time, I now strongly suspect that the number one culprit is bad cooling (besides wobble): for such a low layer height, putting a fan some centimeters far is not enough. I will soon try with a fan duct and let you know.

Edited 2 time(s). Last edit at 11/03/2014 02:54PM by cristian.

Quote
BackEMF
For now i managed to compensate the coupler wobble by turning the rod so that the two wobbles cancel eachother out a bit (it still wobbles a bit)
I went to 2 hardware stores today only to find rods that are uncapable of rolling on a flat surface. :-) Not bend a lot, but bend enough to leave them on the shelf and look further..

At the top, the rod wobbles not more than 2 millimeters, which become 0.1 - 0.2 millimeters in the printed pieces...
What is the diameter of the threaded rod you are looking for? Maybe thicker is straighter?

Quote
cristian

Quote
BackEMF
For now i managed to compensate the coupler wobble by turning the rod so that the two wobbles cancel eachother out a bit (it still wobbles a bit)
I went to 2 hardware stores today only to find rods that are uncapable of rolling on a flat surface. :-) Not bend a lot, but bend enough to leave them on the shelf and look further..

At the top, the rod wobbles not more than 2 millimeters, which become 0.1 - 0.2 millimeters in the printed pieces...
What is the diameter of the threaded rod you are looking for? Maybe thicker is straighter?

From my experience, thicker rods are straighter and also stainless thread is usually better quality.
Holding the top end of the rod in a bearing is a bad idea, because any wobble in a bent rod will be transferred to both the X and Y axis at least some.
The Z rod cannot control this movement completely, unless it is a thick smooth rod.
The advantage of using a thin threaded rod is that it can bend a little itself and take up some of the movement.
Use the best bearing system that you can find (so far Delrin bushes seem to be best for me)

So to have a machine that uses all the features mentioned above;
Smooth Z rods thickest you can use.
Threaded Z rod the thinnest you can use.
No retaining bearing at the top end so the threaded rod can move as needed and not force any unwanted movement to the axis.
Use stainless steel thread for better formed thread.
A good bearing system or use bushes that are free and give a good sliding action.

---

[regpye.com.au]
"Experience is the mother of all knowledge." --Leonardo da Vinci

Quote
regpye
Smooth Z rods thickest you can use.
Threaded Z rod the thinnest you can use.
No retaining bearing at the top end so the threaded rod can move as needed and not force any unwanted movement to the axis.
Use stainless steel thread for better formed thread.
A good bearing system or use bushes that are free and give a good sliding action.

Thanks Reg. Have you ever managed to remove completely the effect of the wobble on the printed pieces in some of your printers? If yes, how?

I would also like to be able to quantify the wobble: I will have a look at cheap microscopes...

Quote
cristian

Quote
regpye
Smooth Z rods thickest you can use.
Threaded Z rod the thinnest you can use.
No retaining bearing at the top end so the threaded rod can move as needed and not force any unwanted movement to the axis.
Use stainless steel thread for better formed thread.
A good bearing system or use bushes that are free and give a good sliding action.

Thanks Reg. Have you ever managed to remove completely the effect of the wobble on the printed pieces in some of your printers? If yes, how?

I would also like to be able to quantify the wobble: I will have a look at cheap microscopes...

I like using Cura, however the slicing doesn't seem to use random starts as well as Sli3r does, so this can give a line mark where the layer begins and ends.
Some of my prints have looked like injection moldings, except for the very top layer that still shows the pattern of the print, but can hardly be felt at all.
I do most prints using 0.3mm layers because it is quicker and strong.
For a good print surface like an ornament I use 0.24mm or 0.2mm layer height.
I have got the best results by using the method described before.
No upper bearing on the threaded rod (a must) Let the threaded rod wobble if it must, and not transfer this wobble to the axis.
Use M5 thread for the Z thread, and use stainless steel with well formed thread. (many cheap threads like zinc coated are very poor threads and make a noise, wobble like hell, and even jam while turning)
Use good bearings or bushes and heavy duty Z rods (I am actually using 8mm on most of my machines, but one has 12mm and has no problems at all)

---

[regpye.com.au]
"Experience is the mother of all knowledge." --Leonardo da Vinci

I try to stay with 5mm, on my first smartrap i have almost no wobble, so it should be possible (also 5mm)

The main reason why i think 5mm is the 'more simple' way to go is because the steppers shaft is also 5mm

Another way to reduce wobble might be one of these: [www.sdp-si.com]
the flexible connection might do the trick. If ithe prices aren't too extreme.

And what about printing a flexible coupling?

cristian,

I really like those z axis rod holders. Can you share?

Here they are: sorry if the code is not very readable...

Quote
BackEMF
Another way to reduce wobble might be one of these: [www.sdp-si.com]
the flexible connection might do the trick. If ithe prices aren't too extreme.

And what about printing a flexible coupling?

I tried to print a flexible coupling and I failed: it is difficult for me to conceive a coupling that preserves rotation and z shift, while being tolerant on x and y shifts...

But if you have some good ideas, please share!

And finally, bye bye wobble!



1cmx1cm cubes, from left to right: 0.1, 0.05 and 0.025 mm layer height.

I obtained this by replacing the Z threaded rod with fishing line in a very naive way, thanks to a motor with perfectly round shaft (I bought it without knowing it was like that, but now it turns out useful...):



Of course there are several disadvantages:

• the cartridge falls when the Z motor has no power, although slowly enough not to be a very serious issue;
• resolution along Z is about 20 times rougher than before, but also this does not seem to be a visible issue: at 1/16 step, it has about 0.006 minimal layer height, or about 0.1 if we consider only full steps, which is more than enough for high quality prints.

I suppose now the horizontal precision limit is given by the gt2 belts. Any ideas about improving this?

Edited 2 time(s). Last edit at 11/10/2014 08:18PM by cristian.

The 4mm aquarium air-tube still worked best for me, as it is rigid enough to do the work and flexible enough in regard to the wobble.
On my first smartrap the M5 rod also seems to be more straight. (i still need to replace the rod on the new one)

Also, i'm wondering about the hotend PID settings you are using and what type/brand of PLA that you used for the grey cubes.

Quote
BackEMF
Also, i'm wondering about the hotend PID settings you are using and what type/brand of PLA that you used for the grey cubes.

I have never changed the PID settings in the firmware I got from Serge's repository.

The PLA is among the cheapest I found: 3DPremium from filimprimante3d. It has some problems sticking to the bed coated with PVA glue, in particular because it warps a little (clearly visible in the above cubes). But besides that, it works great.

Actually, thanks to Reg's hotend I no longer care about what PLA I am using since as long as I don't incinerate the filament, it does not clog. With that grey PLA I used temperatures from 170°C to 220°C, optimizing either bed sticking, oozing, printing speed or whatever other issue I had, depending of course on the purpose of the object I was printing. For example the cubes are printed at 190-205°C at the first layer, and 170°C the rest at low speed (10mm/s) to optimize quality. Moreover, I still need to improve the cooling system.

But it is true that at this level of precision a filament with very slight variations in diameter may affect printing quality. So I will try with other filaments to see if I get better results.

I found that there are many differences in PVA glue per brand, currently I use Bison D2 (in a mixture of 75% water 25% PVA) and that works exellent for me.

Quote
cristian

Quote
BackEMF
Another way to reduce wobble might be one of these: [www.sdp-si.com]
the flexible connection might do the trick. If ithe prices aren't too extreme.

And what about printing a flexible coupling?

I tried to print a flexible coupling and I failed: it is difficult for me to conceive a coupling that preserves rotation and z shift, while being tolerant on x and y shifts...

But if you have some good ideas, please share!

I am using aluminium flexible couplings, but I did something a bit different.
Some people have said that this is wrong, but I am sure they haven't tried it.
Flexible couplings are suppose to accept a motor shaft at one end and the threaded rod at the other. The rod end and the motor shaft should not touch each other because it causes a transmission noise to add marks on your print. There should be a space between the shafts.
What I do is the different part, I drop a 4mm ball bearing in between the motor shaft and the threaded rod.
Why??? The rods are now touching, but the transmission noise does not happen because the ball is touching on the centre part of each motor shaft and the threaded rod. The transmission noise happens when square ends are touching and react against each other. The ball stops this happening and also takes up the movement in one direction, but still allows the shafts to flex as needed.
Removing the movement up and down makes for a better flexible joint that prints more accurately.
It costs almost nothing to try this and if you don't get the results that you want, you can always take the ball out. I leave them in mine because of the better prints I get.

Now I will expect a number of people that "know better" to tell me that I am wrong, but I am sure that they haven't tried this out.

---

[regpye.com.au]
"Experience is the mother of all knowledge." --Leonardo da Vinci

That's an interesting idea that i sure want to try.


looks like that coupling does something similair, but then with a plastic 'spacer'. It would eventually be printable aswell.

And what about setting the Z drive shaft in bearings? this way the shaft cant wobble, but the coupling has to be flexible to compensate for misalignment.

Quote
cristian
the cartridge falls when the Z motor has no power, although slowly enough not to be a very serious issue;

And here is a solution to the Z fall, which also doubles stepper resolution along Z: a couple of hooks that halve the weight of the carriage, enough to avoid movement in the absence of power and external forces.

Quote
regpye
I drop a 4mm ball bearing in between the motor shaft and the threaded rod.
Why??? The rods are now touching, but the transmission noise does not happen because the ball is touching on the centre part of each motor shaft and the threaded rod. The transmission noise happens when square ends are touching and react against each other. The ball stops this happening and also takes up the movement in one direction, but still allows the shafts to flex as needed.

I haven't tried but it makes sense to me.
Still I will do my best to avoid threaded rods as much as possible! You cannot imagine how silent is my printer now: it is hard to hear stepper motors printing if the fan is above 50% - and no, it is not a noisy fan!

why not use GT2 belt?

Quote
BackEMF
why not use GT2 belt?

The only Smartrap design I saw so far with belts on the Z axis is one of Reg's designs, so he can better tell the pros and cons of using it. The reasons why I did not use gt2 belts are that

• I should have changed considerably the design of my printer, in particular I should have used some geared system to increase the step resolution along Z;
• I should have bought belt and pulley;
• I had already fishing line here.

On the other hand, with fishing line no change was necessary, it took about 10 minutes to mount everything after printing the two plastic supports I needed, at nearly zero cost.

Actually I don't understand why people don't try as much as possible to use fishing line instead of belts: it is so convenient and cheap...